【摘要】 构件损伤边缘的识别是影响构件损伤成像分辨率的关键因素之一。为提高构件损伤边缘的成像精度，文章应用激光超声检测技术进行了基于深度学习的构件损伤成像图像优化的研究。在对航空航天用4A01铝板中超声Lamb波进行分析和处理基础上，成功分离出信号中Lamb波的A0模态的幅值。基于A0模态的幅值完成了扫描区域的可视化成像，提出了利用卷积神经网络优化损伤边缘的算法，开展构件损伤边缘优化实验。结果表明，该算法针对各损伤类型实现结构相似性指数（SSIM）平均提升0.064 2，极大地提高了构件损伤边缘区域的成像精度。更多还原

【Abstract】 Identification of component damage edges is one of the key factors affecting the resolution of component damage imaging. To improve the imaging precision of the component damage edges, the deep learning-based image optimization of component damage imaging was studied in this paper while using laser ultrasound detection technology. Based on the analysis and processing of ultrasonic Lamb wave in aerospace4A01 aluminum plate, the amplitude of A0 mode of Lamb wave in the signal was successfully separated. On the basis of amplitude of A0 mode, the visual imaging of the scanned area was completed. An algorithm for optimizing damage edges using convolutional neural network was proposed to carry out the component damage edge optimization experiments. The experimental results show that the algorithm achieves an average improvement of 0.064 2 in the structural similarity index measure(SSIM) for each damage type, which significantly improves the imaging precision of damage edge region of components.更多还原